Name : V.

Hariharasudhan
Register number : 14PHD1049
Department : SENSE
Date of joining : 14-Jul-14
Assignment : RM-assignment unit 4
 Assignment on Optimization algorithms
Objectives:

The objective is to submit an assignment on Optimization algorithms. I have chosen a

research paper of my interest. In this assignment, how the algorithm has been applied in the field
of research with detailed methodology has been explained.

A paper which is used ACO algorithm:

“Mobility-aware Ant Colony Optimization Routing for Vehicular Ad Hoc Networks”

Vehicular Ad hoc Networks (VANETs) are a special type of Mobile Ad hoc Networks
(MANETs), made by vehicles communicating among themselves, and by vehicles
communicating to devices located in the margins of roads and highways. The main characteristic
of a VANET is the high speed of network nodes – that can go up to 200 km/h –, and that impacts
directly on the ability the network has to deliver data, given we might have a network formed for
just a small amount of time. It has been shown in several works that ant-based routing can be
successfully applied to both wired and wireless networks. This work proposes Ant Colony
Optimization (ACO) procedures that take advantage of information available in vehicular
networks – such as the vehicles’ position and speed –, in order to design an ant-based algorithm
that performs well in the dynamics of such networks. The authors have also adapted the Dynamic
MANET On-demand (DYMO) routing protocol to make use of the ACO procedures proposed in
this paper, and the resulting bio inspired protocol, MAR-DYMO, had its performance evaluated
in an urban scenario and compared against a few other routing protocols. The obtained results
suggest that making use of environmental information can make ACO algorithms more suitable
for routing in vehicular ad hoc networks.

Application of Ant-colony optimization algorithm:

Vehicular Ad hoc Networks (VANETs) are a special type of Mobile Ad hoc Networks
(MANETs), made by vehicles communicating among themselves (Vehicle-to-vehicle
communication, V2V), and by vehicles communicating to devices located in the margins of
roads and highways (Vehicle-to infrastructure communication, V2I). The devices installed in the
vehicles are called OBUs (on-board units), and the ones located in the roads, RSUs (roadside
units). In V2V, each OBU works in ad hoc mode, being able to forward messages through
multiple hops, but in this mode the network connectivity is highly dependent on the vehicle’s
density and mobility pattern. In short, the main characteristics of a VANET are the high speed of
network nodes and their reasonably predictable mobility pattern, since the vehicles are supposed
to follow the transit conventions, their trajectory is also supposed to follow the road lines. The
goal of this work is to devise and evaluate bio-inspired procedures that take into account some of
the information available in vehicular networks, such as the vehicles’ position and speed. The
idea is to use this information to help the routing decisions, having, in the end, procedures that
adapt well to VANETs.
In the beginning of the 1990s, the combinatorial optimization based on ant colonies
(ACO, Ant Colony Optimization) emerged as a novel bio-inspired technique intended
to solve hard combinatorial optimization problems. ACO is a metaheuristic , i.e., an approximate
algorithm used for obtaining good enough solutions for hard combinatorial optimization
problems within a reasonable computational time. The source of inspiration for the ACO
metaheuristic was the foraging behavior of real ants. When looking for food, the ants initially
explore the area nearby their nest, in a random way, but leaving – on the path they pass through
– a chemical substance called pheromone This substance is later used by an ant to find the way
back to the nest, as ants are able to follow pheromone tracks. As soon as an ant finds a food
source, it evaluates the source and brings some of the food back to the nest. In the way back to
the nest, it again deposits pheromone on the path it passed by. The amount of pheromone that
was deposited in the path that might depend on the quantity and quality of the food will guide
other ants to the food source. This indirect communication through pheromone trails enables the
ants to find the shortest path between the nest and the food source [9]. This characteristic of the
real ant colonies is exploited in artificial colonies, being the basis for the ACO technique.

The proposal: Mobility – aware ant colony optimization routing

The central component of an ACO algorithm is a parameterized probabilistic model,
which is called the pheromone model , and in this context, two mechanisms are of the most
importance: the way the pheromones are deposited in the paths where the ants pass through and
the way these evaporate within the time. These two mechanisms will be discussed in the
following subsections.

A. Pheromone deposit:
The level of pheromone on a path indicates how good that path is, i.e., it reflects the
quality of the path, its cost. Considering the link `ij , an ant walking on it – going from node i to
node j –, deposits an amount of pheromone on the path. In terms of implementation, when the ant
arrives at node j, the pheromone level of the entry – in the routing table of j having i as both
destination and next hop, is increased of a PR is the expected probability of successfully
receiving a message sent through a given distance, i.e,the probability of j receiving a message
from i or vice versa,given the distance between them is d meters. tlink is the route lifetime, given
by the Kinetic Graphs framework, and tMAX is a value defined as the maximum route lifetime.
This upper bound exists to avoid having too large values for the route lifetime estimation. The
probability of reception is a good indicative of the path quality, but only in cases where the
distance between the sender and the destination nodes is less than the wireless transmission
range. For multi-hop routes whose length is bigger than the transmission range, the only
indicative of quality we have is the estimated lifetime for that path.

B. Evaporation process:
Just like it happens with real ants, an ACO algorithm simulates the evaporation process
the pheromone trails – left by the ants while traveling through the links – experience. From a
practical standpoint, the evaporation process is necessary to avoid a too fast algorithm
convergence towards a suboptimal region. In other words, it is a way to escape from a local
optimum, favoring the exploration of new areas in the search space. The evaporation mechanism,
where is the so-called pheromone evaporation rate, the pheromone level associated with that
path.
In several ACO algorithms, such as ARA , ARAMA and Ant-DYMO, to name a few, the
evaporation rate is fixed and the same for all links, usually a value found in
some empirical way. Since we have available more information about the links than the usual ad
hoc routing protocols – such as, AODV (Ad hoc on-demand distance vector routing) or DYMO
(Dynamic MANET on-demand routing) , we are going to propose different evaporation rates for
every link, based on the assumption they are actually different, and hence, should behave
differently in regard the pheromone evaporation. Through the Kinetic Graphs, we have now an
estimation on the duration of a path, which will be represented by tlink. With that information,
we have an idea on when two nodes will stop being neighbors, i.e., the link between them will
break, and we will set up the evaporation mechanism so that it will completely evaporate after
tlink seconds. In practice, the route will be removed – evaporated – after the link is supposedly
broken, so the idea is to eliminate invalid routes in a more accurate way, exactly when they
become invalid. We will now proceed to calculate the rate needed to evaporate a link within tlink
seconds. Hence, Assuming there are no other interferences in the pheromone level of this
arbitrary link, such as pheromone deposit. As the evaporation mechanism is performed every tev
seconds , we can calculate k, that tells the number of times the evaporation process was
performed.

Application of the proposed technique to an existing routing protocol :

In this section, we will apply the techniques presented in the previous section to the
DYMO routing protocol, and obtain the Mobility-aware Ant Colony Optimization Routing
DYMO, or simply MAR-DYMO, for short.
A. Dynamic Manet On-Demand (DYMO) :
DYMO is a reactive and unicast routing protocol for multihop wireless networks and is
considered a successor of the popular AODV. Its basic operations are the route discovery and
route maintenance.
1) Route Discovery:
The route discovery procedure is performed when a node in the network wants to send
data to another node for which it does not know a route for. At this point, the first node creates a
route request (RREQ) message and floods the network with it. The RREQ messages keep a list
of visited nodes, so that a node, upon receiving a RREQ message, can update its routing table
with a route to the source of this message. After receiving an RREQ message, the node checks
whether it has a route to its destination, and if it does, it sends a route reply (RREP) message
back to the source of the RREQ, telling it about this route. An RREP is also sent if the receiving
node is the actual destination of the RREQ message.
2) Route Maintenance:
The route maintenance procedure takes care of eliminating invalid routes from routing
tables, attributing to each route a duration time. When a route is successfully used in the
forwarding of data packets, the lifetime associated with that route is extended. When the route
lifetime expires, its associated route is removed from the routing table. This way, DYMO only
keeps routes that are being used.

B. Mobility-aware Ant Colony Optimization Routing DYMO (MAR-DYMO)

As we have seen, DYMO is a reactive protocol. We are going to turn it into an ACO
algorithm by adding the pheromone deposit and pheromone evaporation discussed in section IV.
We start by modifying the HELLO message from the DYMO protocol to add the fields.
Now our HELLO messages are not periodic anymore, as it is in DYMO, but they will be sent by
the vehicles when needed, in an aperiodic fashion, managed by the Kinetic Graphs framework.
That way the nodes will keep a table with updated info on their neighbors. This info allows a
vehicle to make predictions on their neighbors, such as their position at a given time instant and
the time they will be still neighbors, for instance. and after each interval ti, a comparison
between the vehicle’s actual and predicted position is performed. If the absolute difference
between the vehicle’s positions is larger than a sufficiently small epsilon – the prediction error
margin –,a new broadcast is performed, i.e., a new HELLO message will be sent. This broadcast
method is used in order to avoid unnecessary broadcasts. When the vehicles itself notices its own
information is outdated, it sends another HELLO message to give the other vehicles current info
for making new and more accurate predictions on its mobility. For turning the protocol into an
ACO algorithm, the routing tables will be adapted to carry, for each route, the pheromone level
associated with it, its evaporation rate – calculated and the predicted lifetime. They are also
changed to work in a multi-path fashion, i.e., they might have more than one route for the same
destination – though with different nodes as the next hop –, and the actual route will be chosen
among the existing ones based on their pheromone levels, by using the roulette-wheel selection
method. Routes with a higher pheromone level will have a bigger probability of being chosen,
but they will not necessarily be the selected ones.